Abstract: Promising materials in a lot of situations come with complicated composition and bond topology. The structural complexity makes such materials difficult to understand and manipulate. To rationale the design of materials with complex structure, data driven computational framework that combines high throughput computation, data mining and machine learning has been established to answer two fundamental questions: How does the atomic scale local structures influence materials properties? How to design accessible chemical reaction pathway to synthesize promising materials in experiments? This talk will demonstrate the capability of my data driven theoretical framework with several case studies in various functional materials system. Additionally, I will propose several future research directions that embrace new opportunities and challenges on accelerating the discovery of new materials with structural complexity.